I know it's unfair compared to a human but I'm more interested in how much it can do. Like what level of leetcode problems can it solve and how well does it use

I know it's unfair compared to a human but I'm more interested in how much it can do. Like what level of leetcode problems can it solve and how well does it use concepts presented in the few shot applications. The whole point is to establish an upper limit on it's generalization power instead of comparing it to a human